Corruption and Economic Growth in Africa
By Miriam School
A dissertation
Submitted to the Nijmegen School of Management Radboud Universiteit
In partial fulfillment of the requirements for the degree of Master of Science
August 2019
MSc Economics
International Economics and Development Supervised by dr. Jeroen Smits
Abstract
Within the existing literature on corruption and economic growth, there are two contrasting viewpoints. On one side, there are scholars arguing that corruption is a heavy constraint on economic growth, especially in developing countries, where corruption often runs rampant. On the other other side, there are scholars arguing that in a second-best world, where weak institutions and inefficient bureaucracy constitute a major impediment to economic growth, corruption might actually act as a trouble-saving device, improving efficiency and increasing economic growth. This thesis uses four different empirical models, estimated using data from 46 African countries, to show that between 2000 and 2017, corruption was negatively
associated with economic growth within Africa. This implies that there is a strong negative correlation between corruption and economic growth and that countries that are more corrupt, tend to grow slower than countries that are less corrupt. Furthermore, the data shows that this effect is weaker in poorer countries.
Index
Abstract ... 3
1. Introduction ... 7
2. Corruption and Economic Growth ... 10
2.1 The relationship between corruption and economic growth ... 10
2.2 Economic growth in Africa ... 12
2.3 Corruption in Africa ... 17
3. Data and Methods ... 25
3.1 Variables and measurement ... 26
3.2 Data ... 28
3.3 Methodology ... 30
4. Analysis and Results ... 32
5. Discussion ... 38
6. Conclusion ... 42
7
1. Introduction
Corruption, from the Latin ‘corruptus’, which simply means ‘corrupted’ (Šumah, 2018), is often defined as the use or abuse of public office for private gains (see for instance Bardhan, 1997; Drury, Krieckhaus and Lusztig, 2006; Gyimah-Brempong, 2002; Odemba, 2010; Wei, 1999; World Bank, 1999). It is a phenomeon that is “related to the beginning of the creation of law and the state” (Šumah , 2018, 63-64) and one that “can be dated with the rise of
mankind” (Rotaru, Bodislav & Georgescu, 2016, 240). The earliest records of corruption date back to the thirteenth century B.C. (Šumah, 2018), in the Persian Empire of the sixth century B.C. the giving and receiving of bribes was regarded as a capital offense, Greek philosopher Plato (born in the fifth century B.C.) proposed capital punishment for public officials who accepted gifts to do their duties, Roman statesman and philosopher Cicero (who lived in the last century B.C.) viewed corruption as one of the most serious crimes (Rotaru et al., 2016). Corruption was widespread during the time of the Spanish Inquisition, when any victim of accusation could make amends when they had enough money, and during the eighteenth century, countries became virtually helpless in the fight against corruption. In 1716, France established a special court which should rule in cases of abuse of royal finances, but those abuses were so extensive that the court was abolished only a year later (Šumah, 2018). Overall, however, while corruption might be an ancient problem (Bardhan, 1997), it has been given more attention in recent times (Šumah, 2018), and in 1997, the World Bank wrote that “global concerns about corruption have intensified in recent years” (1). Research on
corruption, especially on its negative impacts, became more common after 1995, when both countries and international institutions became more aware of the possible risks and problems of high levels of corruption (Šumah, 2018; World Bank, 1997). This does not mean that before the second half of the 1990s nobody wrote about or paid any attention to corruption, but it does mean that the number of research papers written on corruption and its
consequences has increased quite heavily over the last twenty-five years.
Nevertheless, despite this increase in research, there is still a lot either unknown or disputed when it comes to this field of research. Corruption is a pervasive and universal phenomenon that affects almost every culture to differing degrees (Everhart, Martinez-Vazquez & McNab, 2010), but it is not something that can be assessed outside of its geographic and historical
8
context. The causes, impact, and consequences of corruption vary from time to time and from place to place (Rotaru et al., 2016; Šumah, 2018; Warf, 2017; World Bank, 1997). While the general consensus within development economics is that corruption has a negative influence on economic performance and growth (d’Agostino, Dunne & Pieroni, 2016a, 2016b; World Bank, 1997), some researchers state that in certain cases, it is possible corruption might actually improve efficiency and accelerate growth (Bardhan, 1997; Hanousek & Kochanova, 2015). As Leff (1964) puts it, “if the government has erred in its decision, the course made possible by corruption may well be the better one” (11). Overall, the exact relationship
between corruption and economic growth remains unclear, and further research can still add a lot to this field.
Since corruption is one of the several factors that have hindered economic development especially in Africa (Warf, 2017) and Africa is one of the most corrupt regions of the world (Desjardins, 2019; Šumah, 2018), this master thesis is focused on the possible relationship between corruption and economic growth in Africa, in an attempt to answer the following research question:
What is the relationship between corruption and economic growth in Africa, and under what circumstances does this relationship change?
This research question is divided into two sub-questions:
- What is the relationship between corruption and economic growth in Africa? - Under what circumstances is the relationship between corruption and economic
growth either stronger or weaker?
To answer those questions, a combination of existing literature and new empirical research is used, estimating a growth equation using nation-level data on corruption, economic growth, and several nation-specific independent variables from 46 African countries. This research adds to already existing research because there is no clear consensus within the literature about the possible effects of corruption on economic growth, as will be discussed to greater length in following chapter. This chapter also contains a section discussing recent economic developments and economic growth in Africa, both as a whole and in the separate countries, and a section discussing African corruption and its causes. This chapter is followed by a chapter outlining the data set and the models used to empirically investigate the relationship between corruption and economic growth and a chapter presenting the results of the different
9
models employed to analyse this relationship. This thesis ends with a discussion of data, models, and methodology, and a conclusion.
10
2.
Corruption and Economic Growth
2.1 The relationship between corruption and economic growth
As stated in the introduction, corruption is often considered to be a strong constraint on economic growth and development (see for instance d’Agostino et al., 2016a, 2016b; Campos, Dimova and Saleh, 2010, 2015; Gyimah-Brempong, 2002; Hanousek & Kochanova, 2015; Mo, 2000; Wei, 1999). It is a major concern for developing countries because of the detrimental effects it might have on development and welfare. The poorest countries in the world, according to the United Nations Human Development Index, tend to also have low scores on the Corruption Perception Index, and while correlation does not imply causation, this does give rise to the idea that corruption influences economic growth negatively (d’Agostino et al., 2016b).
The relationship between corruption and economic growth has been tested empirically by several authors, including Gyimah-Brempong (2002), Mauro (1995) and Mo (2000). Mauro (1995) uses extensive data sets, focusing on a sample of developing and developed countries and controlling for several variables, to investigate this relationship. He finds that corruption has a significant and negative impact on economic growth: an improvement in the corruption index of one standard deviation is associated with an increase in the average growth rate of GDP of 1.3 percentage point (Mauro, 1995, 701). Gyimah-Brempong (2002) uses data sets from several African countries to estimate a growth equation, and he finds that corruption has a significant negative effect on growth rates, implying that African countries could increase economic performance by reducing corruption. According to his model, developing countries could achieve better economic performance by improving their institutions and reducing corruption than they could through foreign aid and external development assistance (Gyimah-Brempong, 2002, 205-206). Mo (2000) comes to a similar conclusion by introducing an index for the level of corruption into a growth equation: he states that “a one-unit increase in the corruption index reduces the growth rate by 0.545 percentage point” (76), most importantly through its effect on political instability. Two meta-analyses, by Campos et al. (2010) and Ugur (2014), concluded that the main reported effect is indeed negative, and Wei (1999) states that the more corrupt a country, the slower it grows.
11
While many authors accept the general idea that corruption slows down economic growth, the available literature remains divided on the channels through which this works. Gyimah-Brempong (2002) and Mauro (1997) state that corruption slows economic growth by
distorting incentives and market signals leading to the misallocation of resources, especially human talent, into rent-seeking activities. It might get bureaucrats to increase government spending in unproductive sectors that allow them to collect bribes and keep them hidden, affecting not only the level of public expenditure, but also its composition (d’Agostino et al., 2016; Gyimah-Brempong, 2002; Mauro, 1997). Furthermore, corruption serves as an
inefficient tax on those who are forced to pay it, causes a loss of tax revenue, reduces the effectiveness of aid flows through the diversion of funds, increases transaction costs because corrupt acts are conducted in secrecy, reduces the productivity of resources by degrading the quality of those resources, increases uncertainty, distorts the allocation of public services, and reduces private investment (d’Agostino et al., 2016a, 2016b; Everhart et al., 2010; Gyimah-Brempong, 2002; Hanousek & Kochanova, 2015; Mauro, 1995, 1997; Mo, 2000; Warf, 2017). Mo (2000) also argues that corruption reduces economic growth through its effects on human capital and political instability, is strongly negatively associated with the share of private investment, and that it reduces the returns of productive activities. Corruption limits economic development by inhibiting growth in literacy and per capita income and preventing a decrease in child mortality (Everhart et al., 2010), and by causing an increase sociopolitical instability and distrust of the state (Mo, 2000; Warf, 2017). Politically unstable countries tend to devote a larger share of their budget to public administration and defense to address
security matters, leading to reduced investment and growth, while poor economic performance may, in turn, also lead to political unrest (d’Agostino et al., 2016a, 2016b). Generally, scholars agree that corruption has a negative effect of economic growth. At the same time, several authors suggests that corruption might actually increase economic
performance and growth, ‘greasing the wheels’ (Méon & Weill, 2010, 244) instead of putting sand in them (see for instance Bardhan, 1997; Campos, Dimova & Saleh, 2015; Everhart et al., 2010; Hanousek & Kochanova, 2015; Leff, 1964; Leys, 1965; Méon & Weill, 2010; Mo, 2000). Leff (1964) argues that while corruption is morally unacceptable, it can serve as an instrument that helps overcome cumbersome bureaucratic constraints, inefficient provision of public services, and rigid laws. A similar claim is made by Méon and Weill (2010), who state that in a second best world, where weak institutions and inefficient bureaucracy constitute a major impediment to economic growth, corruption might actually act as a trouble-saving
12
device, improving efficiency. They show that corruption is less detrimental to economic growth in countries where the entire institutional framework is weaker. According to Drury et al. (2006), corruption might also be economically beneficial because it tends to favor the most efficient firms, and Hanousek and Kochanova (2015) show that on a firm level, corruption can increase productivity. To be able to compete when high levels of corruption are present, weaker firms must either become more efficient or exit the productive sector (Drury et al., 2006), increasing economic performance. Furthermore, corruption can serve as speed money, and in highly restrictive regulatory environments, corruption can positively influence
economic growth by stimulating entrepreneurship and efficiency (Campos et al., 2015). Méon and Weill (2010) state that while corruption is, on average, associated with more
disappointing economic performance, this “does not prevent the correlation from being positive in those countries where governance is mediocre” (244). Those authors do not argue that corruption is efficient in itself or that countries plagued with a very inefficient
institutional framework, that might benefit from corruption, should allow it to run rampant. Rather, under some circumstances, corruption is more better for the economy than the alternative (Drury et al., 2006; Leff, 1964; Méon & Weill, 2010).
This short overview of existing literature with regards to the relationship between corruption and economic growth shows that there are two distinct viewpoints. The first one, that is more universally accepted, is that that corruption “has few virtues” (Drury et al., 2006, 123) and has a negative impact on economic growth, while the second one is that in some cases, corruption is better than the alternative and might actually accelerate economic growth (Drury et al., 2006; Leff, 1964; Campos et al., 2015). According to Campos et al. (2015), in the end, the body of empirical research on the economic consequences of corruption is not conclusive. 2.2 Economic growth in Africa
Unequal wealth and income distributions, poverty, and corruption are all widespread in many countries in Africa. In 2016, more than half of the total number of the world’s extremely poor lived in sub-Saharan Africa (World Bank, 2017), and many countries in the region are among the poorest in the world. However, according to the African Development Bank (2019), the state of the continent as a whole is actually quite good. After a substantial slowdown in growth in 2015 and 2016 (see figure 2.2.1), the region managed to recover remarkably well (African Development Bank, 2017, 2019).
13
Figure 2.2.1 Real GDP growth in Africa, 2010-2020 (African Development Bank, 2019, 3)
In 2016, real GDP growth in Africa slowed down from about 4 percent in 2015 to only 2.1 percent, a slowdown that was mostly caused by weak global economic growth, especially in China, a continued fall in commodity prices, a drastic drop in oil prices, and several regional shocks including continued armed conflict, epidemics, and droughts (African Development Bank, 2017, 2019). The worldwide recession that hit the continent that year caused a sharp drop in commodity prices and lowered exports, impacting growth rates throughout Africa, especially those of the major commodity exporters (African Development Bank, 2017, 2019; International Monetary Fund, 2019; World Bank, 2019). However, the continent recovered remarkably fast: the following year, Africa’s economy saw an increase in economic growth of 1.5 percentage point, from 2.1 percent in 2016 to 3.6 percent in 2017. Furthermore, GDP growth is projected to accelerate to 4.1 percent in 2019 and 2020, see figure 2.2.1 (African Development Bank, 2017, 2019). Even in 2016, in the middle of an economic slowdown, Africa remained the second fastest-growing region in the world, after developing Asia (African Development Bank, 2017), and apart from China and India, no other emerging or developing economy is expected to see growth percentages as high as Africa in 2019 and 2020 (African Development Bank, 2019).
While improved economic growth across Africa has been broad and far-reaching, there has been and still is considerable variation across and between countries and regions (African Development Bank, 2019; International Monetary Fund, 2019). It is important to remember that all African countries should not be lumped together and viewed as a whole, as there are
14
widespread and important differences between the countries, culturally, socially, geographically and economically, and those should not be ignored or swept aside.
Table 2.2.1 Real GDP growth in Africa, 2010-2020 (African Development Bank, 2019, 3)
As table 2.2.1 shows, there is considerable variation in economic growth across Africa’s different regions. In 2016, for example, East Africa, recorded real GDP growth of 5.1 percent, while West and Central Africa only grew with 0.5 and 0.2 percent, respectively. Those last two regions were dragged down by the economic recession in Nigeria, caused by a persistent fall in oil prices and continuing policy uncertainties (with economic growth contracting to -1.5 percent), and the poor performances of Equatorial Guinea (estimated economic growth of -8.2 percent) and Chad (estimated economic growth of -3.4 percent) (African Development Bank, 2017). In general, East Africa is the continent’s fastest growing region, followed by North Africa, West Africa, Central Africa, and Southern Africa. According to the African Development Bank (2019), East Africa is projected to achieve growth rates of 5.9 percent in 2019, with several countries, including Djibouti, Ethiopia and Tanzania, recording above-average growth rates, while countries like Burundi and Comoros only grow slowly because of considerable political uncertainty. North Africa is projected to achieve growth rates of 4.4 percent in 2019 and to account for 40 percent of Africa’s projected 4 percent growth, but average GDP growth in the region is highly erratic because of the unstable development of Libya, one of the region’s largest economies. Growth in both Central and Southern Africa is gradually recovering but remains below the African average, and growth in Southern Africa is mainly subdued due to South Africa’s weak development and the affect that has on
15
There are many reasons why Africa’s regions differ so much with regards to their economic performance and growth. As table 2.2.1 shows, in 2016, there was a wide gap between GDP growth in oil-exporting and non-oil-exporting countries due to a strong decline in oil prices. This decline heavily affected growth rates of major oil exporters such as Algeria, Angola, Nigeria and Sudan, contributing to an economic recession in Nigeria and impacting economic growth in both West Africa and Africa as a whole. Nigeria is the continent’s largest economy, accounting for a large share of African GDP, but only a small share of the continent’s
economic growth (see figure 2.2.2). The recession in Nigeria had a far larger impact on total GDP growth than the recessions in countries such as Chad, Equatorial Guinea and Libya (African Development Bank, 2017).
Figure 2.2.2 Contribution to GDP growth in Africa, by country, 2010-2020 (African Development Bank, 2019, 9)
Another factor that heavily impacted growth, especially in several countries in North Africa, was the spill-over effect of the Arab Spring, which spread from Tunisia to Egypt and Libya, see figure 2.2.3.
16
Figure 2.2.3 Growth in Arab Spring-affected countries of North Africa, 2013-2018 (African Development Bank, 2017, 26)
The Arab Spring led to political instability and reduced oil production in Libya, severely impacting average growth rates in North Africa (African Development Bank, 2017). After a number of years of considerable economic recession, however, the Libyan economy
rebounded faster than expected. In 2017, the African Development Bank predicted real GDP growth of -3 percent in 2018, but in reality, Libya recorded a growth rate of over 10 percent that year (African Development Bank, 2017, 2019).
Lastly, there are several country-specific drivers and brakes when it comes to economic growth, causing considerable heterogeneity across Africa. West-African countries such as Liberia, Sierra Leone and Guinea faced an economic slowdown caused by the Ebola crisis in 2014, and the disease resurfaced in 2018 in the Democratic Republic of the Congo. Growth in several countries in Central and Southern Africa increased due to higher agricultural output after continued droughts, while Botswana’s growth rates accelerated due to improved diamond trades and Mauritius’s steady growth is mainly driven by tourism. A considerable number of African countries, including but not limited to Burundi, the Democratic Republic of the Congo, Libya, Somalia and South Sudan, continues to experience armed conflict, hindering economic activity and constraining economic growth. This number has decreased over the past decade, but Africa still accounts for the largest share of the world’s armed conflicts (African Development Bank, 2017).
17
Generally, non-oil-dependent countries and non-commodity exporters have recorded
relatively stable growth rates over the past couple of years, while oil-dependent countries and major commodity exporters saw a relatively steep decline in growth rates in the last few years (African Development Bank, 2017, 2019; International Monetary Fund, 2019). It remains important to remember that the aggregate figures discussed earlier in this chapter mask considerable differences across countries, not just with regards to economic performance and growth, but with regards to the drivers of this growth as well.
2.3 Corruption in Africa
Corruption is a universal phenomenon that affects almost every country in the world to differing degrees (Everhart et al., 2010), but according to Šumah (2018), the most corrupt countries in the world are either developing countries or countries in transition. Low-income countries, countries with a closed economy, low media freedom and a relatively low level of education, and countries that are engaged in some form of armed conflict are generally more corrupt. Corruption is not a phenomenon that arises due to one or a few specific factor(s), but Warf (2017) states that corruption generally flourishes in poor countries, where the poor often face demands for bribes to obtain the public services they rely heavily on, in secretive
environments where deals and decisions are made out of the public’s view, in undemocratic societies that lack mechanisms for accountability and the enforcement of laws, in countries with low literacy rates, where people are uninformed and unaware of the actions of their government, in countries without effective independent media, that usually serves as a
watchdog and a whistle blower, and in more patriarchal societies. Corruption does also reflect cultural norms, as it is often seen as a normal part of life and another part of doing business in countries where it is widespread and endemic (Odemba, 2010; Warf, 2017).
In general, corruption is widespread in developing, low-income, and transition countries, including many countries in Africa, but there are significant differences between and within region, see figure 2.3.1 below.
18
Figure 2.3.1 Corruption Perception Index, by country, in 2018 (Transparency International, 2019)
In general, Africa as a whole is widely considered to be among the world’s most corrupt regions. Of the ten most corrupt countries in the world, according to the 2018 Corruption Perception Index by Transparency International, five are African: Somalia, South Sudan, Sudan, Guinea Bissau, and Burundi (Desjardins, 2019; Transparency International, 2019a). The same index shows that sub-Saharan Africa is the most corrupt region in the world, with an average score of only 32 (on a scale of 1 to 100), while the Middle East and North Africa, that includes a total of five African countries, has an average score of 39 and ranks fourth out of the world’s six regions (Desjardins, 2019).
Corruption is a real problem in Africa. It is a highly visible aspect of African politics, and while this is not in any way unique to the continent, African corruption remains pervasive and among the world’s most severe (Odemba, 2010; Warf, 2017). In 2014, ninety percent of Africa’s population, roughly one billion people, lived under very or extremely corrupt governments (Warf, 2017), see table 2.3.1.
19
Table 2.3.1 African corruption by level of severity, 2014 (Warf, 2017, 28)
Warf (2017) states that compared to other parts of the world, African nations are particularly prone to severe corruption. Average incomes are low, a vast part of Africans is illiterate, there is a large number of repressive governments, and many African countries are engaged in some type of armed conflict. Furthermore, corruption in Africa reflects its long and tragic history of colonialism and oppression by the world’s richer countries. Territorial boundaries that were in place before African countries gained their independence remained in place, reinforcing tribal conflicts and civil war, and post-independence political conditions were not helpful in terms of reducing corruption, with foreign aid being misappropriated and institutional adjustment policies foisted on the continent by international organizations impoverishing millions and reducing government resources.
Still, corruption in Africa is not a straightforward or homogenous problem, and there is no such thing as typically ‘African’ corruption (Warf, 2017). As shown in figure 2.3.2 and 2.3.3 below, Africa might be considered the world’s most corrupt region, but the actual level of corruption varies widely from country to country.
20
Figure 2.3.2 Corruption Perception Index, African Union, 2018 (Transparency International, 2019a)
21
In very broad terms, the highest levels of corruption in Africa are observed in countries, including Sudan, South Sudan and Somalia, that are marginally functional and usually unable to provide even minimal government services, wracked as they are by civil war and tribal conflicts. Somalia, with a score of only 10 on the Corruption Perception Index (Desjardins, 2019), has been left without a functioning government for over two decades, while South Sudan (13) and Sudan (16) are both engaged in violent conflict (Warf, 2017) and Libya (17) faces challenges of instability, terrorism, war and conflict caused by the disintegration of the state following the Arab Spring (Transparency International, 2019b). Somalia is the most corrupt country in the world, with only South Sudan and Syria coming close, but on the other side of the spectrum, the Seychelles, with a score of 66, and Botswana, with a score of 61, are the least corrupt African countries, ranking above several European and American countries, including Spain (58), Costa Rica (56) and Greece (45) (Desjardins, 2019). Those countries both have relatively well-functioning democratic and governance systems, that help contribute to their score, but they mostly remain an exception rather than the norm in a region where corruption is generally high (Transparency International, 2019c).
While overall performance in Africa is relatively poor and corruption is high, there are improvements visible all over the continent. Côte d’Ivoire and Senegal are among the most significant improvers in the entire index, moving from 27 points in 2013 to 35 points in 2018 and from 36 points in 2012 to 45 points in 2018, respectively. Those improvements may be caused by the positive consequences of institutional reforms undertaken in both countries and the political will of their leaders in the fight against corruption. Côte d’Ivoire is also, as seen in figure 2.3.4, one of the best performers of the entire continent when it comes to the percentage of citizens that think corruption has not increased in the past year.
22
Figure 2.3.4 Change in corruption in 28 countries in sub-Saharan Africa, 2016 (Transparency International, 2015)
There are also a couple of countries, including Botswana, Burkina Faso, Lesotho and Senegal, where citizens think of corruption as being on the wane in their own country (Transparency International, 2015). In those countries, excluding Burkina Faso, a greater proportion of people state their government is doing well, rather than doing badly in their fight against corruption, see figure 2.3.5 below (Transparency International, 2015).
23
Figure 2.3.5 Government efforts in fighting corruption, rated by their citizens (Transparency International, 2015)
On the other hand, there are also several African countries that experienced sharp drops in their CPI scores, including Congo, Liberia, Ghana and Mozambique. Over the last seven years, Mozambique dropped 8 points, moving from 31 in 2012 to 23 in 2018 (Desjardins, 2019). Furthermore, as figure 2.3.4 shows, in 2015, a majority of Africans thought that corruption in their country had increased over the last twelve months, with especially high numbers in South Africa (83%) and Ghana (76%) (Transparency International, 2015). That same year, 18 governments (out of 28) were seen as completely failing to address corruption.
24
The majority of Africans (up to 64 percent) think their government is doing a poor or very poor job at handling corruption, and only 32 percent said their government was doing fairly or very well at fighting corruption. In a country like Madagascar, up to ninety percent of all citizens say their government is either doing fairly or very badly when it comes to fighting coruption, and in countries like Liberia, Zimbabwe, and Benin, this sentiment is shared by four out of five citizens (Transparency International, 2015).
Overall, figure 2.3.2 shows that the difference between the worst and the best performing African country is 56 points, only 6 points smaller than the difference between the most and the least corrupt European countries (Denmark scores 88, Russia has a score of 26)
(Desjardins, 2019), while figure 2.3.4 shows that while over seventy percent of citizens in Botswana though corruption decreased in their country in 2015, only seventeen percent of citizens in South Africa thought the same thing (Transparency International, 2015). This highlights the large variation between African nations, reinforcing the need to consider their historical, political and economic situations. Corruption must be understood both contextually and geographically, as the emergence of corruption depends on a large number of interrelated factors that vary significantly among countries.
25
3. Data and Methods
It is not yet clear what the relationship between corruption and economic growth really might be, especially when including several other variables that are commonly thought to affect economic growth. It is not enough to rely on existing literature to draw any conclusions about the effects of corruption on economic growth in African countries, and it is necessary to turn to the available data and empirical analysis to shed more light on and make more substantial claims about this relationship.
In existing literature about economic growth there is a large number of factors that are
commonly thought to be associated with or to influence economic growth, including the share of investment in GDP, the rate of population growth, the initial level of real GDP per capita, human capital, technological advancement, maintenance of the rule of law, government consumption, life expectancy, levels of schooling and literacy rates, fertility rates, inflation rates, openness to trade, the level of democracy, distortions and political uncertainty and, of course, corruption (Barro, 1996; Levine & Renelt, 1992; Mauro, 1995; Mo, 2000; Upreti, 2015). It is unfeasible to include all possible variables that are of any importance in a model, because the real effect of many of them is still unknown, there is often a lack of reliable data, especially for developing countries, and a data set with all those variables would simply be too large to work with efficiently. Therefore, the model that will be developed and used here, even though it is based on multiple existing models, will not be exhaustive. This is also not the aim of this model or of this thesis as a whole: the main goal here is not to explain economic growth in itself, but to take a closer look at the relationship between specifically corruption and economic growth in Africa. The variables that are included here are widely believed to have a substantial impact on economic growth and do often appear to explain a large part of the variation in economic growth between countries (Upreti, 2015), but here, their main purpose is to serve as variables that enable us to control for specific, existing differences between countries outside of their level of corruption and their rate of economic growth, not to explain as much of the variance in economic growth as possible. The
relationship between corruption and economic growth remains the main focus, but including other variables makes this analysis more robust.
26 3.1 Variables and measurement
The data on economic growth, GDP per capita, total natural resource rents, export rates, and political stability were taken from the World Bank (2019). The data on average years of schooling, population size, and life expectancy were taken from the Subnational Human Development Index Database (Global Data Lab, 2019). The data on corruption was taken from the Corruption Perception Index (Transparency International, 2019a). Table 3.1.1 below describes each of the variables and their predicted effects on economic growth.
Table 3.1.1 Overview of the dependent and independent variables included in the following models and their predicted effects
Description Predicted Effect
Growth of GDP per capita (% per year) Dependent variable
Score on the Corruption Perception Index (1-100) + / -
Gross Domestic Production per capita in 2000 (1000 US$)
-
Political stability and absence of violence/terrorism (1-5)
+
Export rate (% of GDP) +
Total natural resource rents (% of GDP) + / -
Total population size (millions) +
Life expectancy at birth (years) +
Mean years of education (years) +
The proxy for economic growth is the yearly growth rate of GDP per capita, which is the dependent variable. It is defined as the percentage change in per capita Gross Domestic Production from one year to the next. This variable can be both positive or negative, as countries can experience both positive and negative growth, and a higher growth rate is generally desirable.
The first and main independent variable is corruption, using a countries score on the Corruption Perception Index in a given year as a proxy. This variable can take any value between 1 and 100, and higher scores indicate less corruption. As discussed a couple of times
27
before, it is uncertain whether the sign of the coefficient of this variable will be negative or positive.
The second independent variable is GDP per capita in 2000 (2011 for South Sudan, as it only became a sovereign state in that year), measured in current US dollars. This variable can take any positive value, and a higher value is preferable. The predicted sign of the coefficient of this variable of negative because of the what Mo (2000) refers to as the ‘convergence tendency’(68): the larger the existing knowledge gap between countries, the easier it is for a less developed country to raise its productivity by learning, imitating, and adapting
technology from leading economies. The general hypothesis is that, holding other explanatory variables constant, poor economies grow faster than richer ones (Barro & Sala-i-Martin, 2004; Mo, 2000; Upreti, 2015).
The third independent variable is political stability and absence of violence and terrorism, one of the World Bank’s Worldwide Governance Indicators. This variable can take any value between 1 and 5, and higher scores indicate more political stability and less violence. Alesina et al. (1992) state that “[e]conomic growth and political stability are deeply connected” (1). An unstable political environment causes uncertainty and reduces the speed of investment and economic development, while poor economic performance in turn may also lead to
governance collapse and political unrest (Alesina et al. 1992; Barro, 1991; Mo, 2000). Therefore, the predicted effect of this variable is positive: when political stability rises, economic growth rises as well.
The fourth independent variable is the export rate. This variable is measured as a share of GDP and can take any value between 0 and 100%, with higher values suggesting that a country is more open to trade (Upreti, 2015). International trade enhances the economy of both importing and exporting countries, and even in developing countries, open economies outperform closed economic every year in terms of real GDP growth (Sachs and Warner, 1995; Upreti, 2015). The predicted sign of the coefficient of this variable is positive. The fifth control variable is total natural resource rents as a share of GDP, indiciating the production and use of natural resources produced by the country. This variable can take any value between 1 and 100%, and higher values indicate that a country produces and uses more natural resources. In the literature, there is a certain uncertainty about the effect of the
28
presence of natural resources on economic growth: a country rich in natural resources might be able to benefit from the production and sale of such wealth, but it is also possible for this country to fall into the natural resource trap and fail to profit from its abundant natural
resources. The export of natural resources can lead to an appreciation of a country’s exchange rate, making other exports more expensive, and investing human and physical capital into one specific industry shrinks other industries (Collier, 2006; Sachs & Warner, 1995; Upreti, 2015). Since previous research was inconclusive, the predicted sign of the coefficient is ambiguous.
The sixth independent variable is population size in millions. This value can take any positive value. Larger countries usually enjoy a larger labor force, making labor cheaper due to its immense availability and providing them with the benefits of economies of scale (Thuku, Paul & Almadi, 2013). The predicted sign of the coefficient of this variable is positive.
The seventh independent variable is life expectancy at birth in years. Higher life expectancy generally indicated that a country has a better healthcare system with access to doctors and hospitals (Upreti, 2015), and Acemoglu and Johnson (2006) show that improvement in health conditions may lead to improvement in economic conditions. Improvement in life expectancy foster the accumulation of human capital (Upreti, 2015), and as such, the predicted sign of the coefficient is positive.
The eighth and final independent variable is education, measured as the average years of education a country’s population enjoyed and used a proxy for the level of the human capital stock. According to Mo (2000), the predicted sign of the coefficient of this variable is
positive, because an educated labor force is better at learning, creating, and implementing new technologies, generating a higher rate of productivity growth. Babatunde & Adefabi (2005) state that education “contributes to economic growth by improving health, reducing fertility and possibly by contributing to political stability “ (5), and their analysis shows that a better-educated labor force appears to have a significantly positive impact on economic growth. 3.2 Data
Data were collected for a total of 46 African countries for the time period from 2000 to 2017. Since it is was impossible to collect data for every country and for every year, the average
29
observations per country was 14.1, with a minimum of 4 (for South Sudan, data was only available between 2013 and 2016) and a maximum of 17.
After the data was collected, it was necessary to filter it to get rid of any outliers that would significantly influence any possible insight this data set would provide. Based on various scatter plots of corruption and economic growth and descriptive statistics, ten data points were regarded as such outliers and excluded from the analysis, reaching a final number of
observations of 650. Those outliers included Libya in 2011, when GDP per capita declined with over 60 percent, in 2013, when economic growth reached 121 percent, and in 2014, when economic growth reached -24 percent, Central African Republic in 2013, when GDP per capita declined with over 36 percent, and Botswana in 2009, when economic growth of -9 percent was coupled with a corruption score of 56.
Furthermore, the data was assessed for correlation between the variables and for
multicollinearity. There is no multicollinearity present: all VIF scores were between 1 and 3, with a mean VIF of 2.09.
A Pearson product-moment correlation coefficient was computed to assess the relationships between the different variables, and many of the variables are actually highly correlated, see table 3.2.1.
Table 3.2.1 Correlation matrix, showing Pearson’s r measuring the linear correlation between each of the variables
Corruption GDP
per capita
Political
stability Export rates resource Natural rents
Population
size expectancy Life at birth Mean years of education Corruption — GDP per capita .3869 — Political stability .6391 .3759 — Export rates .1415 .5061 .3168 — Natural resource rents -.4235 .1606 -.2170 .5081 — Population size -.1584 -.1417 -.4472 -.3039 -.0334 — Life expectancy at birth .4059 .4343 .2315 .1238 -.0154 .6247 — Mean years of education .4482 .6135 .3063 .3801 -.0330 .0022 .3499 —
30
There is a considerable positive correlation between education and GDP per capita (r = 0.614, n = 650, p < 0.000), export rates and GDP per capita (r = 0.506, n = 650, p <0.000), export rates and natural resource rents (r = 0.508, n = 650, p < 0.000), and political stability and corruption (r = 0.639, n = 650, p < 0.000). While those high values show that correlation might be a problem, it was decided not to remove any variables. In models like this one, high correlations are to be expected, because many variables that have a significant influence on economic growth, also heavily influence each other. Education, for instance, contributes to economic growth by improving health (Babatunde & Adefabi, 2005), affecting life
expectancy, which in its turn also contributes to economic growth. Therefore, all variables that are explained in the last section are kept in the model.
3.3 Methodology
In order to attempt to answer the research question laid out in the introduction, it is necessary to specify, run, and interpret various different regression models. At first, a simple ordinary least squares regression, solely including economic growth as the dependent and corruption as the only explaining variable was run, to investigate whether or not there is any basis for the idea that corruption is correlated with economic growth, based on this data set. Secondly, a more complex OLS regression was run, including economic growth as the dependent variable and corruption, GDP per capita, political stability, natural resource rents, export rates,
population size, life expectancy, and mean years of education as the independent, explaining variables. This model is a modification of the models used by Barro (1996) and Upreti (2017), and serves to get a clearer picture of the relationship between corruption and economic
growth by controlling for specific differences that exist between countries. To control for those differences even more, both a fixed effects and, after a Hausman test (chi2 = 38.06, p <
0.000), random effects model were estimated. Those models take into account that this research employs panel data, and enable us to capture country-specific effects that occur because different countries have different attributes, starting points, and have to deal with different circumstances.
Those four models do allow us to take a closer look at the relationship between corruption and economic growth within this data set, but ignore possible interaction between corruption and other variables. Lučić, Radišić & Dobromirov (2016) show that there is evidence for
31
capita is lower, meaning that poorer countries might grow faster than richer countries when corruption levels are similar. According to Shabbir, Anwar & Adil (2016), there is some evidence that corruption acts as sand in the wheels in more politically stable countries, while in less politically stable countries, it actually greases the wheels, and Mo (2000) states that the negative effect of corruption on economic growth is larger in politically unstable countries. By adding interaction terms one by one to the OLS regression used as our second model and including the significant interaction terms in a fixed effects and, after a Hausman test (chi2 =
33.92, p < 0.000), a random effects model, we can gain more insights into the specifics of when corruption is detrimental or beneficial to economic growth.
32
4. Analysis and Results
This section presents the results of the different regression models detailed in the previous chapter, especially focussing on the estimated effect of corruption on economic growth, the main focus of this thesis.
At first, descriptive statistics of the different variables used in the models were generated, shown in table 4.1.
Table 4.1 Descriptive statistics, giving mean, st. deviation, minimum and maximum for each variable
Variable Mean Standard
Deviation Minimum Maximum Economic growth 2.141 3.967 -18.491 18.066 Corruption 30.383 10.780 11 65 GDP per capita 1.043 1.318 1.245 7.143 Political stability 1.956 .877 0 3.7 Export rates 32.659 1.984 6.2 94.034 Natural resource rents 14.198 13.310 .001 68.778 Population size 22.551 29.991 .5 76.3 Life expectancy 59.041 7.519 41.7 190.9 Mean years of education 5.011 1.984 1.3 10.1
Over the seventeen years covered in this data set, average growth of GDP per capita in the 46 included countries was 2.14 percent, with a standard deviation of 3.97. Without the outliers that were removed in an earlier stage, the minium of this variable is an economic decline of 18.49 percent, experienced by Zimbabwe in 2008, and its maximum is economic growth of 18.07 percent, also experienced by Zimbabwe, but in 2010. This shows that economic growth does not just vary a lot between countries, but can also be extremely volatile within a country. The mean of the continent’s score on the Corruption Perception Index is 30.4, with a standard error of 10.8, but those scores vary widely, from a minimum of 11, South Sudan in 2016 and Sudan in both 2011 and 2013, and a maximum of 65, Botswana in 2012.
33
Overall, the poorest country in this data set (measured in current US dollars) is Ethiopia, with a GDP per capita of 124.46 dollars in 2000, and the richest country is Libya, with a GDP per capita of 7142.77 dollars in that same year. Mean GDP per capita is a little over 1,000 US dollars, but with a sizable standard deviation of over 1,300 dollars.
The average score with regards to political stability is 1.95 on a scale of 1 to 5, with a standard deviation of 0.88, a minimum of 0 (Central African Republic and South Sudan in 2014 and Sudan from 2009 to 2011), and a maximum of 3.7, Namibia in 2008.
The average export rate (as a share of GDP) is 32.66 and a standard deviation of 1.98. The minimum export rate within this data set is 6.2 percent, Burundi in 2012, and the maximum export rate is 94.03 percent, the Republic of the Congo in 2017.
Mean natural resource rents, as a share of GDP, is 14.2 percent, with a standard deviation of 13.3. The minimum of this variable is only 0.001 percent, Mauritius in 2014 and 2015, and its maximum is 68.8 percent, Libya in 2004.
Mean population is 22.55 million people, with a standard deviation of 29.99. Cape Verde has the smallest population of all included countries, only 500,000, and Nigeria is Africa’s largest country, with a population of 190.9 million.
The mean of life expectancy is 59.04, with a standard deviation of 7.59. Life expectancy at birth was lowest in Sierra Leone in 2003 (41.7 years) and highest in Algeria in 2017 (76.3 years).
The mean of mean years of education within this data set is 5.01, with a standard deviation of 1.98, a maximum of 1.3 (Niger in 2004, 2005 and 2006) and a maximum of 10.1 (South Africa in 2009, 2015, 2016 and 2017).
After inspecting the data, a total of seven different empirical models, as detailed in section 3.3, were estimated. The results of those regressions are presented in table 4.2 below.
34
Table 4.2 Regression outputs, showing beta coefficients and t- and z-values of all variables in the five different models Ordinary least squares Ordinary least squares Random effects Ordinary least squares (including interaction term) Random effect (including interaction term) Constant .9054 (1.95) -.3805 (-.27) -.8008 (-.46) .1000 (0.07) -.1587 (-0.09) Corruption .0407 (2.83)*** .0764 (3.37)*** .0725 (2.71)*** .0370 (1.29) .0254 (0.75) GDP per capita (1000 $) -.3271 (-1.94)* -.4104 (-1.80)* -1.0545 (-2.88)*** -1.3548 (-2.85)*** Political stability .8719 (3.36)*** 1.2456 (4.08)*** .9170 (3.53)*** 1.2354 (4.04)*** Export rates -.0091 (-.67)* -.0093 (-.54) -.0130 (-.94) -.0131 (-.75) Total natural resource rents .0557 (3.30)*** .0791 (3.84)*** .0645 (3.73)*** .0893 (4.24)*** Population size .0257 (4.10)*** .0291 (3.34)*** .0259 (4.16)*** .0291 (3.31)*** Life expectancy -.0199 (-.84) -.2484 (-1.62) -.0084 (-.35) -.0106 (-.33) Mean years of education -.2107 (-1.91)* -.8008 (-1.62) -.2439 (-2.19)** -.2996 (-1.92)* Corruption * GDP per capita .0234 (2.23)** .0310 (2.26)** Absolute values of the t and z statistics are in parentheses. ***, **, and * denotes significance of coefficients at 1%, 5% and 10% respectively
In both ordinary least squares regressions, the first model including only corruption as explaining variable and the second model adding GDP per capita, political stability, export rates, natural resource rents, population size, life expectancy at birth and mean years of education as explanatory variables as well, the coefficient of corruption is positive and highly significant (p = 0.005 in the first and p = 0.001 in the second case). It is important to note here
35
that within the data used, higher scores on the Corruption Perception Index indicate lower corruption, and that this positive coefficient thus points to a negative relationship between corruption and economic growth: overall, more corrupt countries grow slower. This is also the case when taking country-specific effects into account and estimating a random effects model: the coefficient is very similar to the coefficient estimated in the second OLS regression, with a p-value of 0.007.
In all reported models the relationship between initial GDP per capita and economic growth is negative, as predicted: richer countries grow slower than poorer countries, possibly because of the convergence tendency discussed earlier. Political stability is, as expected, highly
negatively correlated with economic growth in every model, indicating that countries that are more stable, grow faster. The relationship between export rates and economic growth is negative, while available literature mainly predicted a positive relationship because more open economies tend to grow faster, but this is only significant, and only at the 10 percent level, in the second OLS regression. This significant relationship disappears completely when a random effects model is estimated. This data set does not provide any evidence of the natural resource curse, as the relationship between natural resource rents and economic growth estimated here is a positive one. Countries that are richer in natural resources, tend to grow faster, not slower as some authors predict. Population size is positively and significantly correlated with economic growth in both the ordinary least squares model and the random effects model, indicating that larger countries (in terms of population) grow faster than
smaller countries. The relationship between life expectancy and growth cannot be interpreted, since it is not significant in any model. According to this data set, mean years of education and economic growth are negatively related, as predicted, but this relationship is not significant when estimating a random effects model. In all other models, countries with a more educated population, that are often richer and more prosperous to start with, grow slower than countries with a less educated population.
The results of those models, excluding interaction terms, show that this specific data set does not provide any evidence of a ‘greasing the wheels’ effect of corruption. While correlation does not imply causation and we cannot say if the observed effects are caused by other, unobserved variables, the data and models used here show that there is a significant negative relationship between corruption and economic growth. This effect is not just observed when estimating an ordinary least squares regression, but also when taking into account differences
36
between countries and allowing for country-specific effects by estimating a random effects model. The different countries included in this analysis are not the same, not by a long shot, but the negative effect of corruption on economic growth still appears to be present.
Because the focus of this thesis is not only on effect of corruption within African countries on economic growth, but also on under which specific circumstances this relationship is smaller or larger, a total of seven interaction variables were generated, one for the interaction between corruption and each other independent variable. Those variables were added to the model one by one and tested for significance, and only one of them actually turned out to be significant in both an ordinary least squares regression and a random effect model: apparently, there is a significant interaction effect between corruption and initial GDP per capita. The coefficent of this new interaction variable is positive, indicating that the negative effect of corruption on economic growth is smaller when initial GDP per capita is lower: in poorer countries, corruption has a less negative effect on GDP growth.
This last result does provide some evidence that corruption might, under certain
circumstances, not be as detrimental to economic growth as thought and as the other models show. Méon & Weill (2010) argue that corruption is on average associated with lower
economic growth, but that this effect seems to decline when moving from rich countries with an efficient institutional framework to poor countries with weak economic institutions. This data set and the models estimate here do not show this, especially because institutional quality is not included in this analysis, but the fact that corruption seems to be more detrimental to economic growth in richer countries and less detrimental to growth in poorer countries does fit this earlier observation. More research would be needed to conclude whether or not the negative effect of corruption does actually disappear in specific countries or under certain, unreported circumstances.
What is important to note here is that corruption seizes to be significantly related with economic growth in all models when this interaction term is added to the model. This makes sense, because the coefficient of corruption should now be interpreted as the unique effect of corruption on economic growth only when initial GDP capita is equal to zero. This is not a feasible scenario, so this coefficient is not as important when interpreting this model.
37
Overall, the models estimated here show that corruption is significantly negatively associated with economic growth. In this vein, it makes sense that corruption is often considered a strong constraint on economic growth in developing countries, where corruption does run rampant and is a lot more prevalent than in more developed countries. At the same time, those results also show that there are still considerable research opportunities in this field: the data and the models used here show that the negative association between corruption and economic growth is smaller in poorer than in richer countries, and it might even be the case that this relationship seizes to be negative in specific countries or under certain circumstances. Further research is needed to test this hypothesis, since those possible specific circumstances are not included in the models estimated here, and including them is beyond the scope of this thesis.
38
5. Discussion
No single research paper is perfect, and this is not even something scholars should necessarily strive for, as it is very likely unattainable. However, the fact that every paper has flaws does not mean that it not necessary to discuss those flaws, and in this last section, the most important of flaws will be discussed.
The first main point of discussion is the problem of defining corruption. The definition of corruption used by many authors and organizations and also used here is that corruption can be defined as the “use of public office for private gain” (see for instance Bardhan, 1997; Drury, Krieckhaus and Lusztig, 2006; Gyimah-Brempong, 2002; Odemba, 2010; Wei, 1999; World Bank, 1999). While this definition is definitely widespread, it is not necessarily a perfect one. Mbaku (2009) argues that by using this definition, one assumes that a central authority exists that exercises certain power over a country, while in countries where the state has collapsed, there is no public office that one might use or abuse. This is a problem in certain African states, including Somalia, which was not included in this research because of a lack of data, which effectively has been left without a functioning government for over two decades (Warf, 2017). It is also possible that in countries where people do not consider themselves members of the nation, even though those states enjoy international recognition, specific groups within such nations actually encourage and welcome exploitation and illegal appropration of specific resources. This is especially the case when central governments and their laws are viewed as external or alien impositions, and while certain practices might be regarded as corruption by outsiders, the people commiting those acts might actually be paraded as heroes (Mbaku, 2009). An aditional problem here is that many legal systems in African countries were inherited from European colonialists, not constructed with the full and effective participation of citizens, and as such, a large proportion of citizens do not accept those laws. Even when national laws prohibit corruption, it may remain pervasive when people do not accept or respect the law, believe that the law and how it is enforced are unfair, and that their government is illegitimate (Mbaku, 2009). In the end, however, we cannot discuss any concept without some form of working definition (Andersson & Heywood, 2009). This definition is definitely not perfect, but we do need a definition to measure corruption, and this one might be as good as any.
39
Defining corruption is not the only problem. We do need a definition to measure corruption, but this is where we also run into problems. Measuring corruption, and especially using the Corruption Perception Index in year-to-year comparisons, is problematic in itself.
The CPI is a composite index that seeks to measure the overall extent of corruption, but it is also an index that does not measures reported cases, prosecutions or proven incidences of corruption, but instead measures perceptions. Andersson & Heywood (2009) argue that this distinction matters, because perceptions influence behaviour and because there can be a striking disjuncture between perception of corruption and personal experience with it. This implies that the method that Transparency International uses to compose their index is
insufficient. One can also argue, however, that this does not necessarily mean that the index is weak, especially if one keeps in mind that this is the way the index is constructed.
Furthermore, this problem is partially adressed by Transparency International’s own Global Corruption Barometer, which also includes personal experience with corruption. The main thing here is that this barometer is a relatively new tool that does not yet cover all that many countries, but in the future, it might (Transparency International, 2015).
Another problem of the CPI is that it is often used as an indicator of improvement of deterioration in countries year by year, especially in cases when the individual score of
countries matters instead of their rank relatively of other countries. Small differences in scores might be caused by variation in data sources between years or by changes in methodology, so one should remain cautious when using the index for year-to-year comparisons (Andersson & Heywood, 2009). On the other hand, the index is very widely used in this kind of
comparisons, not just by academic authors but by Transparency International themselves as well (Thach et al., 2017; Transparency International, 2019a, 2019b). In the end, it is important to remember the index’s shortcomings, but those shortcomings do not imply that the index as a whole should not be used in this kind of research.
Apart from possible issues that surround the use of the Corruption Perception Index, the data and models used in this research are facing the same possible shortcomings that many similar data sets and models in similar economic literature face. As Keele (2008) puts it, “statistical models are always simplifications, and even the most complicated model will be a pale imitation of reality” (1).
40
This dataset is imperfect because it is a cross-country study. As Upreti (2017) states, countries may use different methods to collect virtually the same kind of data, using slightly different definitions. Though this kind of study is imperfect, it is the best available data for a study like this, and data was collected from databases and sources that were as uniform as possible. Even with standard international definitions, country-specific measures may still deviate, and the results presented here are only as reliable as possible.
It is possible and perhaps even likely that this study misses some key indicators of economic growth. In other empiriral literature on the different variables that affect economic growth, factors like literacy rates, technological improvements, government consumption, and institutional effciency have a positive correlation with economic growth (see (Barro, 1996; Levine & Renelt, 1992; Mauro, 1995; Mo, 2000; Upreti, 2015). This study fails to include those variables, often because of the lack of data on those indicators for developing coutntries. Many different indicators were considered, but data could not be found for most of the
countries included in this sample. Those missing variables might have created omitted
variable bias, and by carefully identifying important variables and including those variables in the model, results could be improved (Upreti, 2017). In the end, however, the models
estimated here do not aim to explain as much variance in economic growth as possible, but to explain the relationship between corruption and economic growth as clearly as possible, and including all variables that influence economic growth is not what should be aimed for. There is also a lack of available data for many countries and for many years. In the end, this data set includes 46 out of 54 countries in Africa, with an average of 14 observations for each country. If more consistent and reliable data had been available for more African countries, the results of the models specified here could have been improved, but this proves to be difficult since specific data is often unavailable for developing countries, especially countries that are facing various types of conflict as many in Africa are (Upreti, 2017).
Lastly, economic growth in many parts of the world, but especially in the developed world, is extremely volatile. In Africa, growth rates of over 100 percent are not entirely unprecedented, as are growth rates of minus 60 percent. Especially when research covers a relatively short time period, this makes it a lot more difficult to observe correlations between such a volatile dependent variable and explaining variables that are a lot more stable. Population size, total natural resource rents, export rates, life expectancy, mean years of education, and even
41
corruption often do not change that drastically from year to year, while economic growth might. This implies that results might change when gathering data for a different or a longer time period, because seventeen years might not be long enough for drastic changes to occur in more stable explaining variables.
42
6. Conclusion
Campos et al. (2015) state that “[a]lthough corruption is much more common in poor than in rich countries, it is also clear that it is not restricted to specific regions or levels of economic development” (521). At the same time, the poorest countries in the world tend to also be the countries with low scores on the Corruption Perception Index, and because of its possible detrimental effects with regards to economic performance and growth, corruption remains a major concern for developing countries, including many countries in Africa (d’Agostino et al., 2016a).
The data and models used in this thesis provide yet another empirical basis for those claims and those concerns. Both when regarding the relationship between corruption and economic growth in isolation and when accounting for existing differences between countries that might have a considerable influence on economic growth, corruption appears to have a significant negative effect on economic growth. This implies that African countries that are more corrupt, or at least have a lower score on the Corruption Perception Index, tend to grow at a slower pace. Those results are in line with those of scholars like Mauro (1995), who used a sample of developing and developed countries to come to a similar conclusion, Mo (2000), who
observed a negative effect of corruption on economic growth in developing countries, and Gyimah-Brempong (2002), who also observed this negative relationship in countries in Africa. Overall, those results do provide more evidence for the hypothesis that corruption can be a strong constraint on economic growth and can significantly slow down economic
progress in developing countries.
At the same time, the data also shows that corruption is less detrimental to economic growth in countries where initial GDP per capita is lower. Poorer countries appear to experience a smaller negative effect of corruption on economic growth than richer countries, an effect that might be explained by other, unobserved variables such as institutional efficiency, as
hypothesized by Méon & Weill (2010), but that also might exist on its own. Further research might be needed to test whether or not the association between corruption and economic growth changes from a negative to a positive one in specific countries or in other countries, as several authors have hypothesized. Corruption seems to at least put less sand in the wheels in poorer countries, but it is not yet clear if it actually greases the wheels in some of them.
43
Overall, this analysis shows that economic performance and growth in developing countries in Africa is highly influenced by corruption scores, and that the relationship between corruption and economic growth is a negative one, even more so in richer countries. This implies that countries might grow faster and improve their economic performance by fighting corruption, as already suggested by the World Bank twenty years ago (Wei, 1999). While correlation does not imply causation, the models used here are in no way exhaustive and are, just like any model used in econometric research, a simplified version of reality, increasing efforts to decrease corruption is likely to increase economic growth and performance. Economic growth is not everything, but especially for developing countries, it remains important, and
45
Bibliography
Acemoglu, D. & Johnson, S. (2007) Disease and Development: The Effect of Life Expectancy on Economic Growth. Journal of Political Economy, 115(6), 925-985.
African Development Bank. (2017) African Economic Outlook 2017. African Development Bank. (2019) African Economic Outlook 2019.
Alam, M.S. (1989) Anatomy of Corruption: An Approach to the Political Economy of Underdevelopment. The American Journal of Economics and Sociology, 48(4), 441-456. Alesina, A., Özler, S., Roubini, N. & Swagel, P. (1992) Political Stability and Economic Growth. NBER Working Paper No. 4173.
Andersson, S. & Heywood, P.M. (2009) The Politics of Perception: Use and Abuse of Transparency International's Approach to Measuring Corruption. Political Studies, 57(4), 746-767.
Babatunde, M.A. & Adefabi, R.A. (2005) Long Run Relationship between Education and Economic Growth in Nigeria: Evidence from the Johansen’s Cointegration Approach, presented at the Regional Conference on Education in West Africa: Constraints and Opportunities, Dakar, Senegal, 2005.
Bardhan, P. (1997) Corruption and Development: A Review of Issues. Journal of Economic Literature, 35, 1320-1346.
Barro, R.J. (1991) Economic Growth in a Cross Section of Countries. The Quarterly Journal of Economics, 106(2), 407-443.
Barro, R.J. (1996) Determinants of Economic Growth: A Cross-Country Empirical Study. NBER Working Paper No. 5698.
